Treatment for cocaine dependence is characterized by high rates of relapse, yet the factors influencing the likelihood of relapse are poorly understood. Exposure to cocaine, stress and cocaine-related cues increase cocaine craving, and genetic polymorphisms in the dopamine D4 receptor subtype (DRD4) influence the effects of cues and drug exposure on ratings of craving. However, craving does not robustly predict drug use or relapse. There are currently no data characterizing the interaction between DRD4 polymorphisms, cues and cocaine exposure on actual cocaine taking, i.e., cocaine self-administration. Incorporating measures of relapse into our established laboratory model is an important objective for medications development because models of cocaine self-administration have predictive validity in screening medications for cocaine dependence. Aim 1: Refine our cocaine self-administration procedures to include measures of relapse. The model is guided by hypotheses supported by pilot data: The likelihood of relapse and the quantity of cocaine self-administered following relapse will vary as a function of (1) the cost of cocaine, (2) the presence of contextual cues associated with cocaine-taking, and (3) noncontingent cocaine administration (i.e., 'priming'). Aim 2: Determine the influence of DRD4 polymorphisms on cue- and cocaine-induced relapse. Data with alcohol have demonstrated that individuals heterozygous or homozygous for 7 or more allele repeats (DRD4L) show increased cue- and alcohol-induced craving and greater relapse clinically than those with fewer than 7 allele repeats (DRD4 S). We hypothesize that cocaine-dependent DRD4 L volunteers will show greater cue- and prime-induced relapse compared to the DRD4 S group. Aim 3: Test the effects of modafinil on measures of cocaine relapse as a function of DRD4 polymorphisms. We hypothesize that modafinil will: (1) decrease the effect of both cues and a cocaine prime on the likelihood of relapse compared to placebo, (2) decrease the amount of cocaine self-administered if cocaine use is initiated, and (3) be more effective decreasing cue-and cocaine-induced relapse in the DRD4 L group than the DRD4 S group. [unreadable] [unreadable] [unreadable]